Our Weirdest Dreams Could Be Training Us for Life, New Theory Says

Two people sit in an immersive art installation by Julius Horsthuis on March 5, 2021 in New York City.
Two people sit in an immersive art installation by Julius Horsthuis on March 5, 2021 in New York City.
Photo: David Dee Delgado (Getty Images)

I once dreamt that I lived inside the belly of a Tyrannosaurus rex. Another dream about a haunted refrigerator that kept reappearing in my kitchen despite my repeated attempts to get rid of it makes me chuckle to this very day. And just last night, for reasons beyond my comprehension, I dreamt that I stole a pair of blue cowboy boots so that I could test them out prior to making a purchase.

Advertisement

Dreams can be deeply weird, no doubt about it. So weird, in fact, that we’re tempted to dismiss them as nothing more than quasi-random crap spewed out by our brains while we’re sleeping. Dreams might feel purposeless, but as Tufts University neuroscientist Erik Hoel argues in his new paper, published today in Patterns, there’s a method to this apparent madness. Bizarre and hallucinatory dreams, he says, help us to process and generalize experiences during our waking life, improving our ability to adapt to situations.

“Life is boring sometimes,” said Hoel in a release. “Dreams are there to keep you from becoming too fitted to the model of the world.”

The new theory is now one of a countless number of attempts to explain the purpose of dreams. Indeed, this field is filled with all sorts of hypotheses, including dreams as wish fulfillment, a side-effect of neural pulses, a process to help with the consolidation of long-term memories, threat simulation, among many other ideas. Hoel’s theory is unique in that it borrows from artificial intelligence, using deep neural networks as an analogue for biological brains.

Hoel got the idea while considering the way computers learn. An artificial neural network is fed a dataset for training, but a problem arises when it becomes too familiar with the data. The AI’s world becomes very small, as it assumes the dataset is a complete and true representation of the real world. In reality, the world can be a chaotic, unpredictable, and messy place. This problem is known as “overfitting,” and it “leads to failures in generalization and therefore performance on novel datasets,” according to the paper.

“During training, artificial neural networks are being fitted to the data,” explained Hoel in an email. “If the network is discriminating between cats and dogs, for instance, it might become fixated on some aspects of cats that is particular to those 100 images that make up the data.” For example, the cat photos might’ve been taken during the day, whereas the dog photos were taken at night.

“By adding noise to the images, or blacking out parts of them, you will improve the generalization to new and novel data sets, like images that contain both night and day,” he said. “I’m arguing that the brain probably faces this problem of learning too well, and dreams help give us exposure to the wildly different stimuli that we need to prevent getting fixated on inconsequential aspects of our lives.”

Advertisement

Or, as Hoel writes in his study, “it is the very strangeness of dreams in their divergence from waking experience that gives them their biological function.” And by “hallucinating” out-of-the-box “sensory stimulation every night,” our brains can escape over-generalizations, resulting in improved task performance.

For example, say you’re learning a new game, such as a racing simulator. It’s possible that your performance will peak at some point, but owing to a dream that provided some unique perspectives on the challenge, your performance was able to improve the next day. At least, that’s according to this line of thinking.

Advertisement

This could explain why we dream about repetitive tasks we’ve performed during the day. It’s our brain trying to think out of the box, and to prepare us for when those repetitive tasks might become less familiar.

Deirdre Barrett, a dream researcher at Harvard University and author of Pandemic Dreams, said the new theory “seems congruent with the most bizarre dreams but not with banal, realistic or repetitive ones,” she wrote in an email. Interestingly, Barrett, who wasn’t involved in the new research, believes there’s a basic fallacy in the quest to pinpoint a “function” for dreams.

Advertisement

“We’d never ask, ‘What is the function of waking thought?’ or at least we’d never expect a one-phrase answer. It’s for everything,” she explained. “I suspect the reason there’s no consensus on the function of dreams is that they’re also performing a vast number of psychological and biological tasks.”

“In my view, the theory is not entirely plausible,” said Antti Revonsuo, a professor of cognitive neuroscience at the University of Skövde, Sweden, and at the University of Turku, Finland. “It claims to take ‘the phenomenology of dreaming’ seriously, but then presents a somewhat misleading idea of what dream research has shown dream phenomenology to be like.”

Advertisement

Revonsuo said several sweeping claims in the paper are “not entirely backed up by dream content studies,” and that the theory “only considers some highly selective evidence.” Some evidence, such as the effects of repetitive video games on dreams (like playing Tetris), are only really applicable to “sleep-onset dreams, not to the late-night dreams that most dreaming consist of,” he said.

To which he added: “I find it problematic that the theory seems to be based on somewhat inaccurate generalizations about dream phenomenology rather than on reviewing and summarizing the overall evidence from the relevant dream research literature.”

Advertisement

Revonsuo described the theory as “novel and interesting,” and he hopes it will generate some “directly testable specific predictions about the contents of dreaming under different circumstances or experimental conditions.” But in its present form, he said “the theory is still too general, quite speculative, and difficult to empirically test in any detail.”

Hoel said his new theory is testable, but “given current neuroimaging techniques’’ and the need for people to self-report their dreams, “finding knockdown supporting or contradictory evidence for theories is actually quite difficult.” Future work will be required to “test the hypothesis, both in animal models and humans, and pursue the idea of dream substitutions,” he said. On that last point, Hoel is hoping to explore the concept of “artificial dreams,” in which fictions, such as those portrayed in novels or films, can accomplish a similar function to the real thing.

Advertisement

Hoel’s ultimate hope is that his hypothesis will prompt researchers to “take dreams seriously as having a function in and of themselves,” and not some “epiphenomenon that occurs during some other process.”

The quest to find the true purpose of dreams, should it even exist, continues.

More: Scientists find a way to communicate with dreaming people

Advertisement

Senior staff reporter at Gizmodo specializing in astronomy, space exploration, SETI, archaeology, bioethics, animal intelligence, human enhancement, and risks posed by AI and other advanced tech.

DISCUSSION

ImALeafOnTheWind
ImALeafOnTheWind

Lucid dreamer here. Developed that as a defense mechanism against recurring nightmares when I was a child. Didn’t realize not everyone could do it until I watched Nightmare on Elm Street “Dream Warriors” with friends and was like “yeah just control your nightmare and give yourself powers as usual, duh” - but they were like “what do you mean?”.

Also, when I just started working in IT in the 90s as a bench tech in a computer repair shop - I would run across problems I didn’t immediately know how to solve (web help for that stuff at the time was sparse). I would go home and sleep on it and sometimes actually have dreams where I would work out the fix.

Then I had variations of this dream where I would go through a whole stressful workday in my dreams - then I’d wake up disappointed that I had to go through the whole day again IRL.